Report of the Gravity Survey ConiDiittee. 221 



At least two pendulums just like each other should be used, in 

 order that any accidental derangement of a pendulum may be 

 detected. Sabine said to me in conversation that there ought to 

 be three, as that would enable you, in the event of any derange- 

 ment taking place in course of transit or handling, to tell which 

 pendulum it was that had got altered. If you had only two, 

 and one got slightly deranged, you could only tell which it was 

 by going back to one of the stations where they had been 

 previously used and swinging them afresh. However, I think 

 two only have as a rule been all that have been used in gravity 

 surveys, and I believe that with care in packing, transporting, 

 and handling, such derangements are not likely to occur. 



Before fixing on the form we must answer the question, Is the 

 correction for the resistance of the air going to be determined by 

 calculation or by experiment ? 



(a) If by calculation, we are restricted to forms for which it is 

 possible to effect the calculation. The pendulum might be a 

 plain cylindrical rod, or such a rod with a sphere at the end. In 

 an invax'iable pendulum, soundness of casting would not be of any 

 very great moment, the observations being strictly differential. 

 If a rod be used, I should prefer the ends being made hemi- 

 spherical, or thereabouts. The exact form is of no particular 

 consequence, for for a small portion of the rod near the end the 

 calculation cannot be effected, whether the rod be left plain, or 

 formed into a hemisphere. The calculation for a sphere would 

 not apply to a hemisphere joined on to a cylinder. But the part 

 of the resistance which depends on what is near the end of the 

 rod forms only a small fi'action of the whole, and if we are 

 obliged to have recourse to estimation for that small portion, the 

 uncertainity thence arising can be only very small, since the rod 

 is supposed to be but narrow for its length. The alternative is 

 to adopt the form mentioned by General Walker, a cylindrical 

 rod with a sphere at the end. I do not think there is much to 

 choose between these two forms. I think the latter would keep 

 up its oscillations somewhat longei', and the former would have 

 to be about five feet long (for a seconds' pendulum) which might 

 perhaps be a little inconveniently long. I do not know however 

 that this would be any serious inconvenience. 



